Amazon AWS CEO Matt Garman pushes back against Elon Musk’s space data centers plan

AWS CEO Matt Garman said Amazon’s more than 900 terrestrial data centers will remain the focus for the foreseeable future, citing high transport costs and the absence of permanent structures in space as key obstacles to space-based AI data centers. The comments come after Elon Musk announced a merger of SpaceX and xAI valuing the combined companies at about $1.25 billion and renewed ambitions for massive Starlink launches and eventual space infrastructure, while Amazon continues to pursue a competing Leo satellite constellation with roughly $10 billion earmarked and an FCC extension request to launch 1,600 satellites.

Analysis

Market structure: Near-term, terrestrial hyperscalers (AMZN, MSFT, GOOG) and data‑center ecosystem players (Equinix EQIX, Cisco CSCO for networking) retain pricing power because transport costs keep heavy compute on Earth. Rising AI compute demand implies >15–25% annual capex growth for the next 2–3 years to close GPU capacity gaps, sustaining suppliers of servers, power/cooling and colo real estate. Commodities impact will be concentrated in copper/transformers and power demand (utility capex), with modest upward pressure on corporate BBB credit spreads for capex‑heavy names.

Risk assessment: Tail risk includes rapid SpaceX Starship reusability or a regulatory greenlight that cuts launch cost per kg by >50% within 3 years, enabling meaningful orbital compute experiments and creating stranded terrestrial assets. Shorter horizons (0–12 months) carry low operational risk to hyperscalers; medium (1–3 years) hinge on GPU supply and energy costs; long (3–7 years) could see meaningful tech disruption if launch economics and thermal/maintenance challenges are solved. Hidden dependencies: specialized chips (NVIDIA), power grids, and insurance/legal regimes for orbital infrastructure.

Trade implications: Favor long exposure to established cloud/colo winners: AMZN (AWS) and EQIX for durable demand; tactically overweight NVIDIA (NVDA) to capture persistent GPU shortage via options. Underweight or avoid pure‑play small satellite hardware vendors or speculative space ETFs that assume rapid commercial orbital data centers; these are binary outcomes with high downside if timelines slip. Reprice durations: expect incremental cloud capex to keep tech capex issuance elevated—favor corporate credit over cyclicals if spreads widen >50bp.

Contrarian angles: Consensus underestimates how sticky terrestrial scale economics are: even if launch costs halve, O&M, latency, repair and legal overhead likely keep unit economics worse than on‑Earth for a decade. The market may be undervaluing colo REITs (EQIX) which capture incremental demand without taking launch/space risk. Unintended consequences include orbital congestion and insurance premiums that could keep commercial space economics unfavorable; a single catastrophic orbital incident would reset valuations for space plays.

Key Decisions for Investors

• Establish a 2–3% long position in AMZN (AWS exposure) within 0–3 months, target +20% upside over 12 months, place stop-loss at -12% to limit drawdown while capture continued AI capex tailwinds.
• Add a 1.5–3% long position in EQIX (Equinix) as a 9–12 month trade to capture colo pricing power; target total return 10–15% and yield cushion ~3–4%; trim if data‑center utilization falls >5ppt QoQ.
• Allocate 0.5–1% notional to a 3–6 month NVDA call spread (buy near‑ATM, sell 15–25% OTM) to play persistent GPU tightness; aim for 2–3x return on premium, limit loss to premium paid.
• Reduce/avoid exposure to small‑cap satellite‑hardware vendors and speculative space ETFs (e.g., ARKX‑like vehicles) or short names with >30% revenue guidance tied to LEO by 2026 and cash burn >$50m/yr; target 30–60% downside over 12–24 months.
• Monitor SpaceX Starship cadence and FCC filings over the next 6–18 months; if certified launches achieve >3 successful rapid‑reuse missions and implied launch cost/kg falls >50%, reallocate 2–3% from AMZN/EQIX into public space‑infrastructure names or suppliers that meet strict profitability thresholds.